< Back to previous page

Publication

A system identification procedure for multibody system dynamics

Book Contribution - Book Abstract Conference Contribution

The governing equations of multibody system dynamics are discovered directly from time-series datasets of a given mechanism using an evolutionary symbolic sparse regression approach without any prior knowledge of the respective mechanism. For the discovery of physical models, a dictionary of basis functions is used, a risk of which is that the dictionary is not large enough to discover the physical laws. This risk is mitigated by following an exploration-exploitation strategy to evolutionarily build and validate the function dictionaries, for which a symbolic sparse regression method based on genetic programming is developed. Illustrative examples are finally considered to demonstrate the capability of the presented approach to distilling the governing equations.
Book: Book of Abstracts of the 10th ECCOMAS Thematic Conference on Multibody Dynamics
Pages: 148 - 149
ISBN:9789634218692
Publication year:2021
Accessibility:Open